The Canon EOS ‘R5s’ may be in the hands of testers [CR2]

[Sporgon]

So 90D with 32.5Mp crop sensor would have the same pixel density of a 83 mpixel (32.5 X (1.6)^2 = 32.5 * 2.56) FF sensor.
With this pixel density DLA according to TDP is f/5.2 A 90mpixel camera would have a DLA f/4.8 to f/5 tops. Which means it needs f/4 lenses or up to f/5.6 as the worst case. Maybe an 80mpixel would be a better choice?

Or this is the rumor of 500mm f/2.8 is all about

I think when people like TDP start talking about the theoretical DLA, the practical effects in actual use can be misunderstood, and leaves people thinking that there is more of a problem than there really is. However.............


As the pixels get smaller and smaller diffraction does begin to bite, and I'm about to put up up a tread to justify my comments in this thread earlier about the 5DS at around f/16, which in itself is not a greatly used aperture, but as the same diffraction begins to creep in at more common-or-garden apertures, like f/8 for example, it is going to become as real issue where people are looking to achieve full DOF and the full IQ potential of their very high resolution FF cameras.

It's written all over the web how an ultra high mp DLA camera will still record more detail than a lesser one that is not diffraction limited, people - even - leave - a - space between their words to emphasis their point, and indeed this can be demonstrated by shooting a suitable test target three metres from the camera in good light and viewing the high mp cameras at it's full output size. However get away from studios and test targets and shoot in real "landscape" light and view the image at real, practical output sizes and the picture is not quite so rosy. ( Or more accurately, not so blue ).

 

[Sporgon]

you're raising problems that don't really exist.

you can test this with MTF calculations if you want a high MP camera will still capture a truer image than a low MP camera, every, single, time. It will also resolve more with every single lens you own.

diffraction ONLY appears based upon your level of magnification and observer distance. It's an over-exaggerated problem because people look at monitors at 100%. That's not really .. well, real-life presentation of an image.

No, I am specifically NOT looking at the full size image at 100% and I am specifically referring to real-life presentation of an image. If you do look at 'monitors at 100%' then it is true that the greater output size of the higher mp camera shows a "truer" image even with diffraction over a lower mp one. What I am referring to is viewing am image from the 50 mp 5DS at a realistic output size, say about 42cm / 300dpi long side, so specifically NOT high magnification, when it has suffered from strong diffraction. Here are some examples. My iMac that I use for my photography has died, so these screen grabs are from my 21.5" machine.

The first is the 100% of the full size image, one at f/5, the other at f/16. So this is the comparison where you say above, that you only see it when looking at 100%, and the difference is quite clear to see, as we would expect. All these images are straight ACR conversions from raw with just 80% of 0.3 px sharpening to correct the AA filter.





This was taken on a Manfrotto 058 placed on a hard surface, no wind, no traffic so there is no vibration to give a difference between 1/250th and 1/20th. Live View, 2 sec timer.

I don't have a 20mp full frame camera anymore, but I used a 28mm prime which is same FOV on crop as 45mm on FF, and shot it to give the same DOF as f/16 on the FF shot, so in other words I was using the 19 mp cropped-in 5DS as a crop camera, so f/9. 19 mp at 300 dpi gives an output size of 18" across long size so a pretty normal viewing size. I then downsampled the FF shot taken at f/16 to compare. I hope that this will show up true when uploaded to CR, but the FF f/16 image has lost it's sparkle - clarity, definition, compared with the 'crop' sensor shot at larger aperture. This clearly shows on an 18" 300 dpi print, and that is most salient point.




Below is the FF image shot at f/5 comparing agains the crop one shot at f/9 as a comparison and to see that the 50mp camera downsized to 19mp has quite an advantage in clarity, as we would expect.









So how can you say this "doesn't exist" ? The diffraction at f/16 has reduced the 5DS's IQ against a ( simulation of ) much cheaper, lower resolution camera significantly. This is why I said "falls off a cliff".

Now I can edit the f/16 shot to improve the clarity, contrast and definition of course. Here's a shot below, compared with the unaltered 19 mp crop shot.



I'm not saying that this is a big issue with the 5DS as I rarely want to use f/16 or over, but it does compromise the IQ, and isn't that why we buy a high mp camera, for IQ ? Or is a 50 mp camera only ever supposed to output 36" images ?

But as the FF sensor becomes even more crammed with pixels this "non" issue will creep down the f stop scale, and once it becomes apparent at everyday apertures like f/8, then I think it's going to be a real "issue".

Just for a laugh here's a comparison between the 5DS at f/16 downsized to the same native output as the old G1X, and the same shot from that 'point and shoot' camera.

 

[Michael Clark]

There are different types of professionals with different priorities. What would you choose if you needed a full frame hybrid camera with IBIS that recorded 10bit 4K at least at 60fps?

An actual cinema camera?

 

[Michael Clark]

If you compare the 5DS to 5D it's about a 1k difference. At the price of the R5.. it's a shame that the S version might come out so soon w/o notice (for those that would have been interested in that vs the 5).

Introductory price for the 5Ds/5Ds R ($3,699/$3,899) were only $200/$400 higher than the introductory prices of the 5D Mark III and 5D Mark IV that both released at $3,499 in 2012 and 2016, respectively. The 5D Mark IV price has fallen about $1K in the four and one-half years since it was introduced while the 5Ds/5Ds R price has remained near what it was at introduction in 2014.

 

[Michael Clark]

Wait! What! When did they drastically lower prices due to overstocks????

5D Mark IV at introduction in 2016: $3,499

5D Mark IV in late 2020: $2,499

EF 70-200mm f/2.8 L IS II in 2010: $2,399

EF 70-200mm f/2.8 L IS III in 2018: $2,199

EF 70-200mm f/2.8L IS III in 2020: $1,899

EOS 6D Mark II in 2017: $2,199
EOS 6D Mark II in 2020: $1,399

Compare that to historical prices of previous models that were still close to introductory price when they were replaced by their successor.

 

[Michael Clark]

A 34 mp crop with 10 fps would be a far better option for birding than an R7 in my opinion (Contingent on the autofocus speed and accuracy of course) and with the R series you can set the body to crop to 1.6, so there should be no need to worry about all that extra data slowing things up.

Each line of the sensor that is used in the crop would have to be read all the way from one side to the other, so the data processing savings at the ADC would only be for the 37.5% of the lines that do not need to be read, rather than for the 60.9% of photosites that would not be used.

 

[Michael Clark]

I like you started with 40 MB hard drive, MFM (not RLL which was more expensive)

I remember buying a 1 GB SD and realizing that that was 25 of those drives.

 

[Michael Clark]

actually a full frame camera is better IMO than simply a crop camera.

say you have a 300mm F2.8. at a 34MP APS-C image size you have essentially a 300mm to 480mm lens F2.8 zoom lens.

That is oh I don't know.. probably a 15K lens if Canon ever made it. You essentially have a built in .6x focal reducer every time you slap a lens on a camera. Even better at times for framing as well, because you can switch in between 1x and 1.6x

You may be changed 2K more for a camera body, but you gain far more versatility when it comes to lenses.

Say canon makes / re-makes the 200-400F4L for the RF mount. That becomes a 200mm to 560mm F4L. Pretty freaking nice. You basically turn a 2x 10K lens into a 3x 20K lens for 2K more and still have the premier optical quality of the 2x lens.

While reach is nice, don't forget the versatility that you lose with it as well.

Except that when you crop you demand more from the lens for the same number of lines per image height.

The same lens will always give 1.6X more lines per image height on a FF sensor than on a 1.6X crop sensor.

 

[Michael Clark]

There may be a tilt-shift in my distant future; all sorts of very special things to be done with one...but they run expensive and I can't justify it. I'd probably want to stick with EF on that one just so it can go on everything I own.

Actually I did come up with an interesting (though perhaps silly) idea--an EF --> EF-M adapter with the tilt-shift stuff in it. Someone pointed out that the image circle edges would be to close...which would be true, if you were adapting to an RF mount, but perhaps NOT with an EF-M mount. You could, in principle put any full frame EF lens onto this thing and have at least some play to tilt and shift onto the crop sensor.

Seems like Canon filed a patent for just such an EF → EF-M adapter with tilt and/or shift on the adapter a few years ago.

 

[unfocused]

...Compare that to historical prices of previous models that were still close to introductory price when they were replaced by their successor.

My point was to question the cause of price reductions being traced to "overstocks." That prices change over time isn't in dispute. I am more inclined to believe that price changes reflect market pressures, rather than overstocks.Overstocks implies a failure to accurately predict the demand. I don't think Canon generally makes those kinds of mistakes.

 

[Richard CR]

Except that when you crop you demand more from the lens for the same number of lines per image height.

The same lens will always give 1.6X more lines per image height on a FF sensor than on a 1.6X crop sensor.

maybe i'm misunderstanding you here.

if it's the same lp/mm then it wouldn't make a difference. ie: a 32MP asp-c crop that already exists would be around 80mp full frame. usually the center out resolves the extremities anyways.

 

[Joules]

Each line of the sensor that is used in the crop would have to be read all the way from one side to the other, so the data processing savings at the ADC would only be for the 37.5% of the lines that do not need to be read, rather than for the 60.9% of photosites that would not be used.

Are you sure that's how it works?

The M6 II currently is the only body with a crop mode that boots speed, as far as I'm aware. And it does 32.5 MP * 14 FPS * 14 bit = 6370 ~ 6480 = 18 MP * 30 FPS * 12 bit. So either, the data Bottleneck in this camera is higher than what either mode uses, or they actually are able to reduce the data proportional to the sensor area they are reading.

 

[koenkooi]

Are you sure that's how it works?

The M6 II currently is the only body with a crop mode that boots speed, as far as I'm aware. And it does 32.5 MP * 14 FPS * 14 bit = 6370 ~ 6480 = 18 MP * 30 FPS * 12 bit. So either, the data Bottleneck in this camera is higher than what either mode uses, or they actually are able to reduce the data proportional to the sensor area they are reading.

IIRC you can't read single pixels in a CMOS, only complete rows. So a 1.6x crop would speed up reads by 1.6x. Processing after that could cut off the sides and only work on the reduced area.

 

[Michael Clark]

maybe i'm misunderstanding you here.

if it's the same lp/mm then it wouldn't make a difference. ie: a 32MP asp-c crop that already exists would be around 80mp full frame. usually the center out resolves the extremities anyways.

A FF sensor has a greater image height than a crop sensor. Thus, the same lp/mm gives a higher number of lines per image height when the full sensor height of each is used.

When you crop, either by using a smaller sensor or by cropping the output of a larger sensor and then display an image at the same size as the uncropped image is displayed, you enlarge the size of the image projected onto the sensor by a greater factor. Thus you enlarge the size of the smallest line pairs that can be resolved by the lens. So the optical quality of the cropped image from the same lens will not be equal to the optical quality of the uncropped image when both are viewed at the same display size. This is because when you enlarge by a higher factor you enlarge everything more, including blur.

 

[Michael Clark]

Are you sure that's how it works?

The M6 II currently is the only body with a crop mode that boots speed, as far as I'm aware. And it does 32.5 MP * 14 FPS * 14 bit = 6370 ~ 6480 = 18 MP * 30 FPS * 12 bit. So either, the data Bottleneck in this camera is higher than what either mode uses, or they actually are able to reduce the data proportional to the sensor area they are reading.

The total data from each line must be read and go through ADC. From that point the parts not needed can be discarded before other processing operations, such as demosaicing, WB corrections, NR, sharpening, etc. are done. It all depends on where the bottleneck(s) are in the camera's path from the sensor to the memory card.

The reduction from 14-bit to 12-bit at higher frame rates with multiple cameras (R5, R6, 1D X Mark III, M6 Mark II) would indicate that at least one bottleneck is at the ADC.

 

[Richard CR]

A FF sensor has a greater image height than a crop sensor. Thus, the same lp/mm gives a higher number of lines per image height when the full sensor height of each is used.

When you crop, either by using a smaller sensor or by cropping the output of a larger sensor and then display an image at the same size as the uncropped image is displayed, you enlarge the size of the image projected onto the sensor by a greater factor. Thus you enlarge the size of the smallest line pairs that can be resolved by the lens. So the optical quality of the cropped image from the same lens will not be equal to the optical quality of the uncropped image when both are viewed at the same display size. This is because when you enlarge by a higher factor you enlarge everything more, including blur.

i see what you are doing.

but in essence, you have this reversed from what I was stating.

if you take a R7 with a 32MP sensor and outputting at 30x20" using a 400mm lens, versus a full frame 83MP sensor outputting at 30x20. The R7 will give you an equivalent focal of 640mm at 32MP.

The worst an 83MP full frame camera can do is the same as 32MP crop camera, and from 400mm to 639mm it's going to resolve more than the R7 would.

 

[Michael Clark]

i see what you are doing.

but in essence, you have this reversed from what I was stating.

if you take a R7 with a 32MP sensor and outputting at 30x20" using a 400mm lens, versus a full frame 83MP sensor outputting at 30x20. The R7 will give you an equivalent focal of 640mm at 32MP.

The worst an 83MP full frame camera can do is the same as 32MP crop camera, and from 400mm to 639mm it's going to resolve more than the R7 would.

If all one has is the 400mm lens then cropping the 82MP R5s to 32 MP does not give "full frame" quality that is better than the "crop body quality" of the 32MP R7. That is what your original comment above seemed to be implying.

If one uses a 400mm lens on the R7 and a 640mm lens on an R5s to get the same angle of view from the same camera position, and if both lenses can resolve the same number of lines per millimeter as projected onto the sensor, then the R5s will have 1.6X more lines of resolution per image height than the R7. But that requires a longer and presumably more expensive lens that can resolve the same number of lines per millimeter as the shorter focal length lens.

If one is concerned with putting the maximum number of pixels on the subject with the same lens from the same shooting position, then the size of the sensor has no effect on image quality. It's the pixel density of the sensor that gives higher or lower quality. If the pixel density of two differently sized sensors is the same, then there's no difference assuming the subject is small enough to fit within the frame of the smaller sensor and both sensors are from the same generation of technology. There's no "higher quality" from the FF sensor in that scenario.

 

[Richard CR]

If all one has is the 400mm lens then cropping the 82MP R5s to 32 MP does not give "full frame" quality that is better than the "crop body quality" of the 32MP R7. That is what your original comment above seemed to be implying.

If one uses a 400mm lens on the R7 and a 640mm lens on an R5s to get the same angle of view from the same camera position, and if both lenses can resolve the same number of lines per millimeter as projected onto the sensor, then the R5s will have 1.6X more lines of resolution per image height than the R7. But that requires a longer and presumably more expensive lens that can resolve the same number of lines per millimeter as the shorter focal length lens.

you are totally missing the point and making this way to complicated. I'm talking about ONE lens here.

A R5s (of 83mp) has the inherent benefit of cropping 83MP down to 32MP (1.6x) crop which effectively gives a "ZOOM" of 400mm to 640mm.

An R7 can only see the field of view equivalent of 640mm with the same 400mm lens. Since it can't go wider (not a full frame sensor) it can't give you a FOV of 400mm. Ever.

In conclusion ..

If you have a Canon EF 400mm F2.8L IS USM;

• A R5s of 83mm will allow you to shoot down to a 1.6x crop of 32MP image size to effectively give you a FOV of 640mm - allowing essentially a 400-640mm F2.8-4.0 zoom. At no point is the lp/mm or lp/image height less than an R7. It's impossible.
• An R7 will allow you to use that same 400mm F2.8 lens to get you an equivalent reach of 640mm F4. and that's it.

So when people quibble about the price of a R5s versus an R7 - they arent taking into account the abilities that a Full frame high MP sensor allows - because a Canon RF 400-640mm F2.8-4.0 wouldn't be cheap, or light. It would be an insane lens that would probably cost in excess of 20K.

That was my point. People that complain about a 4K USD camera body instead of a 2K camera body aren't thinking about all the possibilities.

 

[KeithBreazeal]

I can't get any of my programs to open the R5 raw files. My Lightroom 6 is up to date but lacking the raw support. Running Windows 7. Any ideas or suggestions?
Thanks.

 

[AlanF]

I can't get any of my programs to open the R5 raw files. My Lightroom 6 is up to date but lacking the raw support. Running Windows 7. Any ideas or suggestions?
Thanks.

Download Adobe DNG converter. It's free, will convert R5 to .dng, which you can open in your LR.